*BSD News Article 30804


Return to BSD News archive

Xref: sserve comp.os.linux.misc:15487 comp.os.386bsd.misc:2478 comp.unix.unixware:5502 comp.unix.solaris:16610
Newsgroups: comp.os.linux.misc,comp.os.386bsd.misc,comp.unix.unixware,comp.unix.solaris
Path: sserve!newshost.anu.edu.au!munnari.oz.au!spool.mu.edu!howland.reston.ans.net!EU.net!uknet!cf-cm!cybaswan!iiitac
From: iiitac@uk.ac.swan.pyr (Alan Cox)
Subject: Re: A good NFS server ?
Message-ID: <1994May19.115606.11109@uk.ac.swan.pyr>
Organization: Swansea University College
References: <CpC9Fq.I2n@acsu.buffalo.edu> <2rcdee$c8@wea.eel.ufl.edu> <Cq1HGz.8w1@hippo.ru.ac.za>
Date: Thu, 19 May 1994 11:56:06 GMT
Lines: 20

In article <Cq1HGz.8w1@hippo.ru.ac.za> csgr@cs.ru.ac.za writes:
>I can't say that I have any experience with Linux as a NFS server, but
>I have also seen rather poor performance with a Linux system running as
>a NFS client.  (About 90K/sec for a Linux client, as opposed to 600K/sec
>for a FreeBSD client, running the same hardware.)
>
>What I did notice, is that Linux does not appear to have a nfsiod.  Is
>this correct?  Maybe someone on one of the Linux groups has some
>suggestions on how to increase performance?
>
>As far as running NFS on FreeBSD goes, that has recently been added to
>the FAQ, and people having questions may want to have a look there.

The problem with Linux as an NFS client is partially poor caching but mostly
the fact that *BSD (and SUN) kernels are hopelessly inefficient at doing
synchronous I/O in blocks under a page size. The 1.1.13 client will be doing
8K NFS and from my tests with a SUN its a heck of a lot better tho by
no means perfect.

Alan